From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 3F438C54EE9 for ; Thu, 8 Sep 2022 21:53:02 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229572AbiIHVxB (ORCPT ); Thu, 8 Sep 2022 17:53:01 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:37846 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229476AbiIHVxA (ORCPT ); Thu, 8 Sep 2022 17:53:00 -0400 Received: from mx0b-00082601.pphosted.com (mx0b-00082601.pphosted.com [67.231.153.30]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 7FA55FE4A4 for ; Thu, 8 Sep 2022 14:52:59 -0700 (PDT) Received: from pps.filterd (m0148460.ppops.net [127.0.0.1]) by mx0a-00082601.pphosted.com (8.17.1.5/8.17.1.5) with ESMTP id 288LqrZB028275 for ; Thu, 8 Sep 2022 14:52:58 -0700 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=fb.com; h=from : to : cc : subject : date : message-id : content-type : content-transfer-encoding : mime-version; s=facebook; bh=0z2cAUmv0R4fHZps4ghap9epLVNFTHABSccNsUhIaAc=; b=mxwTsBLpj5YfYHnpfDq1cncpH50BnQnJHQRD4p3I5zXhlZP8DdW7X0ondQ32r3IRFkxD TKMHzKS7XYKqcpj1xon45Fg7dKceJvgbHtx29+Y9weo7tqQNQU7stmERP99rFqFZLttC DklPy4RzV/frRq9uftTM9Ksab/8Xs1a/w0s= Received: from maileast.thefacebook.com ([163.114.130.16]) by mx0a-00082601.pphosted.com (PPS) with ESMTPS id 3jfk74jrqv-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT) for ; Thu, 08 Sep 2022 14:52:58 -0700 Received: from twshared7509.08.ash8.facebook.com (2620:10d:c0a8:1b::d) by mail.thefacebook.com (2620:10d:c0a8:82::e) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31; Thu, 8 Sep 2022 14:51:39 -0700 Received: by devbig007.nao1.facebook.com (Postfix, from userid 544533) id A94DC8601C58; Thu, 8 Sep 2022 14:51:35 -0700 (PDT) From: Keith Busch To: , CC: Keith Busch Subject: [PATCHv4] sbitmap: fix batched wait_cnt accounting Date: Thu, 8 Sep 2022 14:51:32 -0700 Message-ID: <20220908215132.3243008-1-kbusch@fb.com> X-Mailer: git-send-email 2.30.2 X-FB-Internal: Safe Content-Type: text/plain X-Proofpoint-ORIG-GUID: aGxvn9wauCA76eXdYLbyZw55b0hZI5fq X-Proofpoint-GUID: aGxvn9wauCA76eXdYLbyZw55b0hZI5fq Content-Transfer-Encoding: quoted-printable X-Proofpoint-UnRewURL: 0 URL was un-rewritten MIME-Version: 1.0 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.205,Aquarius:18.0.895,Hydra:6.0.528,FMLib:17.11.122.1 definitions=2022-09-08_12,2022-09-08_01,2022-06-22_01 Precedence: bulk List-ID: X-Mailing-List: linux-block@vger.kernel.org From: Keith Busch Batched completions can clear multiple bits, but we're only decrementing the wait_cnt by one each time. This can cause waiters to never be woken, stalling IO. Use the batched count instead. Link: https://bugzilla.kernel.org/show_bug.cgi?id=3D215679 Signed-off-by: Keith Busch --- v3->v4: Merged to the latest block for-6.1/block that contains the dependency fix that I attempted to implement in v3. block/blk-mq-tag.c | 2 +- include/linux/sbitmap.h | 3 ++- lib/sbitmap.c | 34 ++++++++++++++++++---------------- 3 files changed, 21 insertions(+), 18 deletions(-) diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c index 8e3b36d1cb57..9eb968e14d31 100644 --- a/block/blk-mq-tag.c +++ b/block/blk-mq-tag.c @@ -196,7 +196,7 @@ unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *d= ata) * other allocations on previous queue won't be starved. */ if (bt !=3D bt_prev) - sbitmap_queue_wake_up(bt_prev); + sbitmap_queue_wake_up(bt_prev, 1); =20 ws =3D bt_wait_ptr(bt, data->hctx); } while (1); diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h index 8f5a86e210b9..4d2d5205ab58 100644 --- a/include/linux/sbitmap.h +++ b/include/linux/sbitmap.h @@ -575,8 +575,9 @@ void sbitmap_queue_wake_all(struct sbitmap_queue *sbq); * sbitmap_queue_wake_up() - Wake up some of waiters in one waitqueue * on a &struct sbitmap_queue. * @sbq: Bitmap queue to wake up. + * @nr: Number of bits cleared. */ -void sbitmap_queue_wake_up(struct sbitmap_queue *sbq); +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq, int nr); =20 /** * sbitmap_queue_show() - Dump &struct sbitmap_queue information to a &str= uct diff --git a/lib/sbitmap.c b/lib/sbitmap.c index cbfd2e677d87..cd631d2e4737 100644 --- a/lib/sbitmap.c +++ b/lib/sbitmap.c @@ -599,24 +599,27 @@ static struct sbq_wait_state *sbq_wake_ptr(struct sbi= tmap_queue *sbq) return NULL; } =20 -static bool __sbq_wake_up(struct sbitmap_queue *sbq) +static bool __sbq_wake_up(struct sbitmap_queue *sbq, int nr) { struct sbq_wait_state *ws; unsigned int wake_batch; - int wait_cnt; + int wait_cnt, cur; bool ret; =20 ws =3D sbq_wake_ptr(sbq); - if (!ws) + if (!ws || nr <=3D 0) return false; =20 - wait_cnt =3D atomic_dec_return(&ws->wait_cnt); - /* - * For concurrent callers of this, callers should call this function - * again to wakeup a new batch on a different 'ws'. - */ - if (wait_cnt < 0) - return true; + cur =3D atomic_read(&ws->wait_cnt); + do { + /* + * For concurrent callers of this, callers should call this + * function again to wakeup a new batch on a different 'ws'. + */ + if (cur <=3D 0) + return true; + wait_cnt =3D cur - nr; + } while (!atomic_try_cmpxchg(&ws->wait_cnt, &cur, wait_cnt)); =20 /* * If we decremented queue without waiters, retry to avoid lost @@ -638,7 +641,7 @@ static bool __sbq_wake_up(struct sbitmap_queue *sbq) * Wake up first in case that concurrent callers decrease wait_cnt * while waitqueue is empty. */ - wake_up_nr(&ws->wait, wake_batch); + wake_up_nr(&ws->wait, wake_batch - wait_cnt); =20 /* * Pairs with the memory barrier in sbitmap_queue_resize() to @@ -663,12 +666,11 @@ static bool __sbq_wake_up(struct sbitmap_queue *sbq) return ret; } =20 -void sbitmap_queue_wake_up(struct sbitmap_queue *sbq) +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq, int nr) { - while (__sbq_wake_up(sbq)) + while (__sbq_wake_up(sbq, nr)) ; } -EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up); =20 static inline void sbitmap_update_cpu_hint(struct sbitmap *sb, int cpu, in= t tag) { @@ -705,7 +707,7 @@ void sbitmap_queue_clear_batch(struct sbitmap_queue *sb= q, int offset, atomic_long_andnot(mask, (atomic_long_t *) addr); =20 smp_mb__after_atomic(); - sbitmap_queue_wake_up(sbq); + sbitmap_queue_wake_up(sbq, nr_tags); sbitmap_update_cpu_hint(&sbq->sb, raw_smp_processor_id(), tags[nr_tags - 1] - offset); } @@ -733,7 +735,7 @@ void sbitmap_queue_clear(struct sbitmap_queue *sbq, uns= igned int nr, * waiter. See the comment on waitqueue_active(). */ smp_mb__after_atomic(); - sbitmap_queue_wake_up(sbq); + sbitmap_queue_wake_up(sbq, 1); sbitmap_update_cpu_hint(&sbq->sb, cpu, nr); } EXPORT_SYMBOL_GPL(sbitmap_queue_clear); --=20 2.30.2